Temporal contrast analysis in human vision
Nine experiments are reported which investigate the mechanisms mediating the perception of temporal contrast, the change of luminance over time, by the human visual system. A distinction is made between linear systems, whose performance may be appropriately characterised in the frequency domain, and nonlinear systems which are sensitive to the temporal structure of visual stimuli. Of particular interest is the role played by systems which are selectively sensitive to relatively slow, nonperiodic luminance changes in the perception of both periodic and nonperiodic temporal contrast. Using a trapezoidal waveform in which amplitude, frequency, and luminance gradient may be varied independently, it is lust shown that the visual system is sensitive to the slope of a temporal profile. An adaptation technique is then used to show that waveform - specific threshold elevation occurs at low flicker frequencies, but not above about 5 Hz. The pattern of results suggests that separate channels exist for fast and slow luminance transitions, irrespective of their periodicity, and that the 'slow' system is subdivided into brightening and darkening channels. In the final two experiments the fast and slow systems are studied in detail by probing the internal representations of step and ramp stimuli, using a masking technique. The step response is consistent with a system whose output is the second derivative of the smoothed input function. The same system can also account for the measured ramp response. This latter finding is likely to be an artifact of the measurement technique, using a 'fast' pulse stimulus. Polarity - specific interactions are observed which indicate polarity selectivity at the level of the 2nd derivative, or the filter output. The experimental findings are integrated within a model of the complete temporal contrast system. While both the structure and the parameters of the fast system are well specified, further work remains on the characteristics of the slow system. The place of temporal contrast analysis within a larger model of movement perception is discussed.